Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.

Poll

Do you rather prefer the particular frequency scale for graphic equalizer frequency bands?

Logarithmic/octave bands
[ 1 ] (50%)
Psychoacoustic frequency scale (Mel, Bark, or equivalent rectangular bandwidth)
[ 1 ] (50%)
Other (please specify the frequency scale in post/reply)
[ 0 ] (0%)

Total Members Voted: 2

Topic: FIR-based graphic equalizer with psychoacoustic frequency scale (Read 421 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

FIR-based graphic equalizer with psychoacoustic frequency scale

Because I'm curious whether or not psychoacoustic frequency scale is better than traditional logarithmic scale for graphic equalizer DSP effect, here's a project about the FIR-based graphic EQ based on perceptual frequency scales like Mel/Bark/ERB and this is an example of using a custom FFT for anything other than audio analysis, in this case, generating the impulse response from frequency-domain function (BTW, it uses cosine-shaped/Hann function filters that are symmetric on the selected frequency scale)

Of course, this project also have a feature to export the IR to a short floating-point WAV file for use with convolvers and the samplerate for the exported IR file can be set (allowing exotic values like 16kHz, 65536Hz as well as not having to set the device samplerate just for exporting)

X

Rooms for improvements:
  • Minimum-phase mode: Yes, I aware of the minimum-phase filters that is not IIR but rather FIR and in fact I've tried to implement this (in-order to have "zero-latency" and no preringing artifacts) but so far, I've failed to properly do real cepstrum method to turn into zero-latency FIR, so anyone can help me to guide towards a minimum-phase FIR filter (while having the exact same frequency response as linear-phase counterpart)?
  • A feature to adjust the frequency band's gain just by clicking on the graph would be appreciated
  • A pre/post spectrum (or oscilloscope since this project also have waveform graph mode) graph, kinda like FabFilter Pro-Q 3 and a feature to disable the analyzer display for both waveform and spectrum graph. I should mention that since all linear-phase filters introduces latency, the pre-spectrum should be latency-compensated unless the FIR filter mode is minimum-phase (which right now is not implemented yet)

Re: FIR-based graphic equalizer with psychoacoustic frequency scale

Reply #1
Why?  IMO it's only useful in hearing laboratories, when comparing one person's hearing frequency response against the standard model of human hearing (or some similar study).

The ideal response of an audio system from microphone through to speakers is "flat".  That way the listener hears exactly what they would if they were present in the original recording environment instead of (or as well as) the microphone.  The human psycho-acoustic response defines how the listener perceives the sound whether live or recorded.

Graphic equalisers have two jobs: to correct for inadequacies in the frequency response on the reproduction equipment (particularly speakers and room acoustics), and to apply artificial listener preferences (potentially to compensate for individual hearing defects) – but be sure that beyond flattening the equipment response, any further compensations are not what you would hear if present at the original recording, in the concert hall or whatever.

Therefore, if working with an inherently (or calibrated) flat reproduction chain, you should be interested in frequency band dB relative to normal perception... and normal perception is already built in!

Making the system frequency response match the psycho-acoustic curve would not be normal perception, high and low frequencies would be artificially boosted and produce and unrealistic colouration (that's what the "loudness" button does).  That might be your listening preference, but it's not what you would hear in real life.

It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: FIR-based graphic equalizer with psychoacoustic frequency scale

Reply #2
Why?  IMO it's only useful in hearing laboratories, when comparing one person's hearing frequency response against the standard model of human hearing (or some similar study).
Well, I'm just curious about a graphic EQ based on frequency scale that matches the auditory system (where it is approximately linear scale at lower frequencies and anything above is approximately logarithmic) or in other words, this graphic equalizer's frequency bands are based on critical bands rather than in octaves as in standard graphic equalizers (including 31-band GEQ)

The ideal response of an audio system from microphone through to speakers is "flat".  That way the listener hears exactly what they would if they were present in the original recording environment instead of (or as well as) the microphone.  The human psycho-acoustic response defines how the listener perceives the sound whether live or recorded.

Graphic equalisers have two jobs: to correct for inadequacies in the frequency response on the reproduction equipment (particularly speakers and room acoustics), and to apply artificial listener preferences (potentially to compensate for individual hearing defects) – but be sure that beyond flattening the equipment response, any further compensations are not what you would hear if present at the original recording, in the concert hall or whatever.

Therefore, if working with an inherently (or calibrated) flat reproduction chain, you should be interested in frequency band dB relative to normal perception... and normal perception is already built in!

Making the system frequency response match the psycho-acoustic curve would not be normal perception, high and low frequencies would be artificially boosted and produce and unrealistic colouration (that's what the "loudness" button does).  That might be your listening preference, but it's not what you would hear in real life.
But what I'm talking about is the frequency bands for graphic EQ being spaced in perceptual frequency scale, since you can pretty much replicate psychoacoustic frequency response using any equalizer effect (including foobar2000's built-in EQ effect) regardless of the frequency scale used for spacing the EQ bands (in fact, most of them have logarithmic frequency spacing) and it is for an another topic, which is related to loudness control in which changing the volume changes the frequency response to compensate for lower volume having less perceived bass and treble

Re: FIR-based graphic equalizer with psychoacoustic frequency scale

Reply #3
But what I'm talking about is the frequency bands for graphic EQ being spaced in perceptual frequency scale
"perceptual frequency" is logarithmic so far as I know...
It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: FIR-based graphic equalizer with psychoacoustic frequency scale

Reply #4
But what I'm talking about is the frequency bands for graphic EQ being spaced in perceptual frequency scale
"perceptual frequency" is logarithmic so far as I know...
While it is true, it is not exactly logarithmic, so here's the image of a spectrum analyzer with Bark scale vs. a standard logarithmic scale:
X
X
So, did you see the perceptual frequency scale being approximately linear on lower frequencies and logarithmic on everything else on the top image?